Cerebras develops wafer-scale AI hardware and integrated software for high-performance model training and inference. Its Wafer-Scale Engine powers the CS-3 system and Cerebras Inference Cloud, which supports OpenAI API-compatible workflows and deployment across on-premise, cloud, and edge environments. The company’s platform is designed for applications including model serving, fine-tuning, pre-training, real-time AI agents, genomics, drug discovery, cybersecurity, and enterprise search. Cerebras brings together specialized chips, systems, and software for organizations building and operating demanding artificial intelligence workloads, with a team of 501–1,000 employees.
ML Algorithm Mapping and Performance Engineer, Core ML
Map machine learning training and inference algorithms to Cerebras AI chip architecture. Model performance bottlenecks, benchmark systems, and inform hardware-software co-design.
Description
- Build analytical and empirical performance models for advanced ML training and inference algorithms
- Assess asymptotic behavior and algorithmic trade-offs across model size, sequence length, batch size, parallelism, and hardware scale
- Develop Pareto frontiers spanning model quality, latency, throughput, memory footprint, communication, and compute cost
- Create prototype implementations and benchmarks for the Cerebras WSE alongside relevant GPU and software baselines
- Investigate system behavior to locate kernel, compiler, runtime, communication, and algorithmic bottlenecks
- Assess emerging methods such as parallel token generation, diffusion and speculative decoding, attention, sparsity, mixture-of-experts, low-precision computation, and distributed training
- Collaborate with research, kernel, compiler, runtime, inference, and architecture teams on implementation and co-design priorities
- Create tools and visualizations to project performance, analyze measurements, and compare design trade-offs
- Present conclusions, assumptions, limitations, and recommendations through technical reports, presentations, and design reviews
- Shape Core ML research priorities, implementation choices, and future hardware and software capabilities
Requirements
- Bachelor’s, master’s, doctoral degree, or equivalent practical experience in computer science, computer engineering, electrical engineering, mathematics, or a related discipline
- Strong knowledge of computer architecture, parallel computing, and systems performance
- Solid understanding of machine learning fundamentals and ML systems
- Experience with analytical performance modeling, algorithmic complexity analysis, benchmarking, or system simulation
- Demonstrated analytical reasoning and problem-solving ability
- Proficiency in Python and working knowledge of C++
- Experience profiling and debugging performance in ML, HPC, CPU, GPU, or accelerator-based systems
- Ability to connect mathematical analysis, experimental validation, and practical engineering recommendations
- Preferred: experience with roofline analysis, CPU or GPU simulators, kernel optimization, or hardware-software co-design
- Preferred: familiarity with CUDA, Triton, PyTorch, JAX, or open-source LLM training and inference systems
- Preferred: knowledge of transformer internals, attention variants, KV-cache strategies, model parallelism, sparsity, quantization, and parallel generation
- Preferred: research publications, patents, or substantial open-source contributions in ML systems, computer architecture, or computational efficiency
- Preferred: experience assessing technology choices for future hardware or software architectures
Benefits
- Opportunities to publish and open-source advanced AI research
- Work with one of the world’s fastest AI supercomputers
- Startup energy combined with job stability
- Straightforward, non-corporate culture that respects individual beliefs
- Ongoing opportunities for learning, development, and support
- Commitment to an equal and diverse workplace